A real-life scenario: We’ve been told to find ways to use AI for our docs

Your organisation’s leadership team have instructed every part of the organisation to define their AI strategy. The edict is vague, and it’s not really clear what you should be trying to achieve with it – other than find ways of doing more work, faster. 

Common causes 

Anecdotes about how AI has reduced cost, saved time, or improved quality are everywhere. Somewhere in there, there’s some real data about this too. Naturally, the leadership team expect their teams to make use of this technology too. But they don’t always have a good understanding of what can be achieved with it around documentation, so the request can seem vague. 

Solution 

While AI is the current focus, from a technical communications perspective we just see this as the latest iteration of something that every team should be considering anyway: how to optimise tools, processes and workflows to make it as efficient as possible to create appropriate quality documentation.  

We’ve always introduced automations here wherever possible, so the AI focus is just increased impetus for this process. But of course, AI has taken on an almost totemic status very quickly: expectations are high, but experience in reality is still catching up. 

1. Find out what’s behind the push for AI.  

As with any transformation initiative, success depends on aligning with business goals. Some of the ways AI can help are: cost savings, reduced time to market, cutting out sticky parts of a process that cause friction between teams, or improving quality. Which of these does your leadership team really care about? 

2. Decide whether to look for strategic improvements or tactical ones. 

Creating good documentation is a complex process, but fundamentally it comes down to getting knowledge out of the heads of various individuals, and into the hands of people who need it. Those people – and the systems they use – are spread across various parts of the organisation (and beyond!), and each have their own particular concerns and priorities. 

To identify places to automate, you can take a strategic approach – consider opportunities that rethink the entire process, tooling and people involved. For example, if the development team are introducing Spec-driven development, can the same specs used to drive coding can also be used as the input to generate documentation against pre-defined templates? Or maybe your support chatbot to draw directly on a central repository of source content – removing the need to create knowledge-base articles as a separate step. Perhaps you don’t need to make information to tell users how to do things with your software product; instead, you focus on helping them express what they want to do, and the software can take care of making it happen. 

This type of strategic solution is likely to have a greater impact ultimately … but of course requires involvement and agreement from multiple stakeholders, so it’s the type of initiative that often stalls because of the complexity. But if the entire organisation is genuinely getting behind thinking through how to make use of AI, then this is the approach to take. 

Alternatively, you can take a more tactical approach: for each step in the process, and each area of your content operation, what are the opportunities for automation and AI? For example, automating documentation QA, or automating maintenance activities by identifying which content is affected by the latest product changes, and drafting the updates. 

As with any process improvements, deciding between strategic and tactical approaches is going to come down to balancing ambition for what can be achieved against the complexity of agreeing and implementing those changes across different parts of the organisation. 

3. Look across your content operation, and identify where AI would have the most impact. 

Your content operation includes all the tooling, processes and activities around creating and delivering your documentation. This can go right from how you collect information about the product through to how your end-users access it, and also includes supporting activities such as governance and managing your tech stack. 

If you don’t have your content operation defined, now might be a good time for that: it will help identify where best to focus. 

With enough effort – and risk-mitigation – AI can be applied to pretty much any part of your content operation. Depending on whether you take a strategic or tactical approach, you might be able to remove entire steps … or just achieve improvements within some of the activities.  

Some factors to help with decisions about where to focus are: 

• Could AI be part of the solution for any of the challenges you’re already trying to solve? For example, if you know people have trouble finding the relevant information in the current documentation, perhaps a Copilot-style AI search is part of the solution. Or if you never find out what’s happened in a product update until after it’s released, perhaps an agent could help with that. 
• How frequently does something need to be done, and how much effort does it take? Setting up and optimising agents or prompts takes time – so the benefit needs to be worth that effort. 
• What’s happening in adjacent teams? Are there initiatives in the teams you work with where you could combine approaches? For example, if you’re responsible for implementation documentation, are the training team already working with product experts to automate some steps of their content development process? Or if you make software user documentation, are your developers designing a new way to enable users to interact with in via the UI? 
• What access to LLMs will you have (and what limitations)? There are token limitations to be considered, as well as the cost of using the LLM or buying AI add-ons for your existing tools. 
• Consider risk. Relying on AI to generate user documentation might be fine for corporate productivity software, but the risks of introducing inaccuracy in a user manual for a medical device are very different, and would need mitigation that might lead you to decide AI won’t add enough value for this activity. 

Special mention: making docs AI-ready is part getting the wider organisation ready to use AI.

If the organisation generally is looking at how AI can be transformative, it’s quickly going to become clear that product documentation that’s ready to be consumed by AI is important. Accurate, well-structured source documentation is essential for other AI systems that depend on the documentation, such as support chatbots or configuration agents. This may mean the best place to focus effort is on validating, refactoring or updating existing content. 

Outcome 

The impact of AI depends on the organisation’s vision and ability to implement change. Aligning your product documentation and content operation to your organisation’s business goals gets the support of the leadership team. As a bonus, it might even take away some of your day-to-day documentation pain points.